A computational analysis of biological behavior: interactions of gaze control and visual attention

Understanding biological behavior through computational modeling is an important and a challenging research topic which requires interdisciplinary corporation. Biological systems are fronted with similar problems computer vision aims to solve, namely, the control over visual sensors and the selective acquisition of information. In this thesis, interactions between gaze control and visual attention in barm owls were investigated. Barn owls make an ideal subjects since they lack any eye movements. Thus, analysis is reduced only to head movements. A novel 3D head tracking system was developed to acquire high accuracy measurements of freely behaving subjects (humans and barn owls). Analysis of barn owls head motion profiles has suggested three main categories of behavior: fixations (no motion), saccades (fast rotational movement to redirect gaze) and peering (pure translational movements). Saccade movements share similarities to the way humans move their eyes. In addition, hypothesis regarding the functional role of peering movements have lead to the development of a novel camouflage breaking algorithm. A better insight to the role of fixations was obtained through additional experiments using a tiny wireless camera, attached to the barn owl’s head. Video sequences recorded from the owl’s point of view represent a close approximation to the visual environment as seen by the owl. These sequences were automatically analyzed and have reveled striking similarities in the allocation of visual attention. 23 T ec hn io n C om pu te r Sc ie nc e D ep ar tm en t M .S c. T he si s M SC -2 00 717 2 00 7